APPFL
Argonne National LaboratoryPrivacy-preserving federated learning framework from Argonne National Laboratory. It simulates synchronous and asynchronous algorithms on HPC systems with MPI, and runs real federations over gRPC or Globus Compute with client authentication, with global and local differential privacy.
The GitHub account `APPFL` is the project's own, not Argonne's organization account.
Openness
5 high confidence- license
- MIT(OSI
- source
- public(the whole framework)
- core features withheld
- no — a DOE-funded national-laboratory project with no paid edition
APPFL is MIT-licensed and developed at Argonne with Department of Energy funding. There is no commercial party positioned to hold features back, and no paid edition is named.
- https://raw.githubusercontent.com/APPFL/APPFL/HEAD/LICENSE recorded 2026-09-27
The LICENSE body: verbatim MIT text, "Copyright (c) 2023 Argonne National Laboratory".
- https://raw.githubusercontent.com/APPFL/APPFL/HEAD/README.md recorded 2026-09-27
README: "This material is based upon work supported by the U.S. Department of Energy, Office of Science, under contract number DE-AC02-06CH11357." No paid tier is named.
- https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/APPFL%2FAPPFL recorded 2026-09-27
Repository record for APPFL/APPFL: license mit, not archived, not a fork, pushed 2026-09-21, 184 stars.
- https://ungh.cc/repos/APPFL/APPFL/files/main recorded 2026-09-27
The 900-file tree of APPFL/APPFL at `main`: one root LICENSE, and no ee/, enterprise/, commercial/, proprietary/ or premium/ path anywhere in it. Path absence only, paired with the statement cited beside it.
Adoption
1 medium confidencePyPI downloads of `appfl`, which carries both the server and the client.
- https://pypistats.org/api/packages/appfl/recent recorded 2026-09-27
last_month = 438 downloads for the package `appfl`.
Capability
4 medium confidenceClients authenticate over gRPC or through Globus, and training can apply differential privacy, so APPFL runs real cross-institution federations as Flower does. Its HPC support is MPI simulation rather than a launcher for production sites, which keeps it below NVIDIA FLARE.
- https://raw.githubusercontent.com/APPFL/APPFL/HEAD/README.md recorded 2026-09-27
README: "APPFL supports MPI for single-machine/cluster simulation, and gRPC and Globus Compute with authenticator for secure distributed training"; "APPFL supports several global/local differential privacy schemes".
- https://ungh.cc/repos/APPFL/APPFL/files/main recorded 2026-09-27
The tree ships launch_server_auth and launch_client_auth notebooks, a gRPC authentication guide and an SSL example.
Verified 2026-09-27